Classificação de sentenças de Juizado Especial Cível utilizando aprendizado de máquina

Source code for the paper publishe in Revista Democracia Digital e Governo Eletrônico.

Link for the paper: https://buscalegis.ufsc.br/revistas/index.php/observatoriodoegov/article/view/316

Project Structure

Classical ML Experiments

In this subproject, we add the pipeline used for text classification using Orange 3. The file can be opened as a project in the Orange 3 tool.

We used the following classical ML techniques:

• k-Nearest Neighbors
• Random Forest
• Naïve Bayes
• feed-forward Neural Networks
• Logistic Regression
• Support Vector Machine
• Neural Network

Deep Learning Experiments

In this subproject, we present the code of the pipeline for text classification implemented in the Python language and open-source tools, such as, NLTK, Scikit-Learn, and Pandas.

The following DL techniques were used:

• LSTM
• CNN (based on Kim (2014))
• Bi-LSTM with Self-Attention Chalkidis, Kampas (2019)

Notes

• We did not make our datasets for text classification available due to the existing personal data in the documents.